Analysis of speech features in female depression patients with anhedonia symptoms
10.3760/cma.j.cn371468-20221122-00707
- VernacularTitle:女性快感缺失型抑郁症患者的语音特征分析
- Author:
Rongxun LIU
1
;
Ning WANG
;
Yang WANG
;
Sanqiao YAO
;
Guangjun JI
;
Shisen QIN
;
Fengyi LIU
;
Zhongguo ZHANG
;
Yange WEI
;
Xizhe ZHANG
;
Rongxin ZHU
;
Fei WANG
Author Information
1. 新乡医学院心理学院,新乡 453003
- Keywords:
Speech feature;
Depression;
Female;
Anhedonia;
Mel-frequency cepstral coefficients
- From:
Chinese Journal of Behavioral Medicine and Brain Science
2023;32(10):901-908
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To explore the speech features of female patients with anhedonic depression and their recognition of pleasure deficient symptoms.Methods:A total of 102 female depression patients who were hospitalized at Nanjing Brain Hospital from September 2020 to October 2021 were selected, including 62 anhedonic depression patients (anhedonic group) and 40 non-anhedonic depression patients (non-anhedonic group). A total of 50 female healthy controls were recruited during the same period.All participants were evaluated by the 17-item Hamilton depression scale (HAMD-17), Snaith-Hamilton pleasure scale (SHAPS), and the temporal experience of pleasure scale (TEPS), as well as voice acquisition.SPSS 23.0 software was used for data processing.Statistical analysis was conducted using one-way ANOVA, non-parametric tests, Logistic regression, and receiver operating characteristic curve.Results:Compared with the non-anhedonic group, the anhedonic group showed significant changes in 15 voice features(all P<0.05), including Mel-frequency cepstral coefficients, formant frequencies, intensity, and energy features.Among these features, Mel-frequency cepstral coefficients exhibited the highest accuracy in identifying anhedonic depression, with sensitivity of 47.5%, specificity of 91.9%, area under curve (AUC) of 0.751, 95% CI=0.686-0.866.Formant frequencies could identify female anhedonic depression, with a sensitivity of 90.0%, a specificity of 40.3%, an AUC of 0.647, and 95% CI=0.605-0.824.Energy features could identify anhedonic deficient depression, with a sensitivity of 60.0%, a specificity of 74.2%, an AUC of 0.679, and 95% CI=0.587-0.804.Intensity features could identify female anhedonic depression, with a sensitivity of 70.0%, a specificity of 58.1%, an AUC of 0.640, and 95% CI=0.554-0.769. Conclusion:Mel-frequency cepstral coefficients, formant frequencies, intensity features, and energy features may have specific changes in female patients with anhedonic depression.The Mel-frequency cepstral coefficients has the highest recognition accuracy for anhedonic symptoms in female depression patients, and is expected to become an objective evaluation index for female anhedonic depression.